WO2001092890A1 - Method for the analysis of picomole amounts of carbohydrates - Google Patents

Abstract

The present invention relates to a miniaturized method to analyse carbohydrates that are present in picomole amounts in a sample. More particularly, the present invention relates to the fluorescent or spectroscopic labelling of carbohydrates, the efficient separation of the labelling reagent from the labelled carbohydrates and subsequent electrophoretic separation for the analysis of the carbohydrates.

Description

METHOD FOR THE ANALYSIS OF PICOMOLE AMOUNTS OF CARBOHYDRATES

Field of the invention

The present invention relates to a miniaturized method to analyse carbohydrates that are present in less than picomole amounts in a sample. More particularly, the present invention relates to the fluorescent labelling of carbohydrates, the efficient separation of the labelling reagent from the labelled carbohydrates and subsequent electrophoretic separation for the analysis of the carbohydrates.

Background of the invention

The recent focus in biological research on the role of glycoproteins in the structure and function of living cells has lead to an increase in the effort to analyse carbohydrates in trace quantities. Mono- and polysaccharide analytical applications in which very low detection limits are desirable, include carbohydrate and glycoprotein sequencing methods, the discovery and identification of new carbohydrates and drug development. Protein-linked carbohydrates can be analysed using a variety of high-resolution techniques, such as high performance liquid chromatography (HPLC)⁸, capillary electrophoresis (CE)9 and mass spectrometry (MS)¹⁰, especially matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) MS. When using MALDI-TOF-MS to analyse carbohydrates highly efficient desalting procedures are required which can complicate trace analysis. Despite this, good quality MALDI-TOF spectra have been obtained from N-glycans derived from submicrogram amounts of tPA by for example

Papac and colleagues⁴. However, a caveat relating to MALDI-TOF-MS of glycans is that quantitation of the oligosaccharides is still controversial and that isobaric (regio- and stereoisomeric) structures, so prevalent in the field of carbohydrates, cannot be determined. For ultrasensitive detection in chromatographic and electrophoretic analytical schemes, it is necessary to derivatize the glycans with a fluorophore, which can be tailored for each specific application. However, in order to obtain quantitave derivatisation, it is necessary to use a large excess of fluorescent tag. This not only requires high purity of the fluorescent tag preparation, but also implies that this excess of tag must be efficiently removed if trace amounts of derivatised glycans are to be detected. In the art this problem is circumvented by derivatising large amounts of glycans and diluting the derivatisation mixture prior to analysis^,! 1. This may result in high theoretical sensitivities, but is not useful for real-world high sensitivity glycosylation analysis, where the amounts of starting material are sometimes only available in less than picomole amounts. It is clear from the above that high performance glycan analysis is very complex and time consuming and that research concerning protein-linked glycans is a difficult challenge. Furthermore there are no easy electrophoretic, miniaturized methods available for glycan-analysis. US patent 5,569,366 describes the use of APTS (9-aminopyrene-1 ,4,6-trisulfonic acid) to derivatize carbohydrates and after separation by capillary electrophoresis picomole amounts of carbohydrates can be detected. However, in said patent, and also in US patent 5,964,999, it is not possible to start with less than nanomole amounts of not yet derivatized glycoconjugates present in a sample. In all prior art concerning APTS usage, a large amount of glycoconjugate (for example a glycoprotein) is deglycosylated, the total sample of released glycans is labelled and then diluted in order to avoid overloading of the analytical separation system by the non-reacted APTS. The present invention relates to a novel, miniaturized, electrophoretic methodology. Furthermore, in this invention the efficient removal of APTS, after derivatizing the carbohydrates, leads to the unexpected finding that samples comprising less than picomole amounts of carbohydrates can be analysed.

Aims of the invention

The present invention aims at providing an identification method for the analysis of carbohydrates. In particular the invention aims at providing an extremely sensitive method for the analysis of underivatized carbohydrates which are present in less than 1 picomole in a sample. The invention further aims at providing a method comprising derivatizing carbohydrates with a fluorescent label, removing the not reacted portion of said fluorescent label, electrophoretically separating said derivatized carbohydrates and detecting said derivatized carbohydrates. The invention further aims at providing an essential clean-up step for the removal of the not reacted portion of the fluorescent label. Said clean-up step is an essential feature of this invention and may comprise the use of a gel filtration and/or gel permeation resin and /or a size-selective membrane. The invention further aims at providing a device for the detection of less than 1 picomole of underivatized carbohydrates present in a sample. The current invention further aims at providing a method comprising a DNA-sequencer. The invention thus provides a sensitive method for sequencing free carbohydrates or carbohydrates derived from glycoconjugates present in a sample. The invention also aims at providing diagnostic tools for the analysis of carbohydrate structures in specific disease processes and for the identification of specific carbohydrate structures derived from recombinant glycoproteins. Finally, the invention aims at providing a sensitive analysis method for carbohydrate structures wherein said carbohydrates are bound to other biomolecules or are derived from organisms.

Figure legends Figure 1 :

Outline of the integrated 96-well sample preparation scheme used. First, the glycoprotein is deglycosylated using PNGase F after binding to the Immobilon-P membrane at the bottom of a well in a 96-well plate. Following release, the deglycosylation mixture is split in two parts. One part is treated with acetic acid in order for full conversion to reducing carbohydrates to occur. Subsequently, the mixture is loaded on a AG-50-WX8 microcolumn in a well of a filterplate. The eluate is loaded on a MALDI-TOF-MS target and subjected to MALDI-TOF mass spectrometry. The other half of the deglycsolation mixture is evaporated to dryness at the bottom of a well of a tapered-96-well plate. The glycans present in the pellets are then derivatized with APTS. The derivatisation is stopped by the addition of water and the mixture is loaded on a Sephadex G10 microcolumn in a well of a filterplate. The eluate is concentrated by vacuum evaporation and loaded on a lane of a sequencing gel after addition of internal standard and formamide. Optionally, both in the mass spectrometry as in the derivatisation-electrophoresis approach, final analysis can be preceded by digestion of the analytes with exoglycosidase arrays.

Figure 2:

Detector response curve shows the linear relationship between the amount of starting glycan and the fluorescence response detected by the ABI 377 detection system over at least three orders of magnitude (from low femtomole to picomole amounts).

Figure 3:

This figure illustrates the use of the disclosed methodology in profiling and sequencing of a standard mixture of triantennary trisialylated N-glycans. 2 pmole total carbohydrate was used as the starting material and after APTS-derivatisation and cleanup using Sephadex G10 microcolumns, the sample was divided in six parts, each containing about 300 fmole of total labelled carbohydrate. 1 part was used to give the native profile (as shown in panel 2). Notice the isomeric resolution obtained. 4 other parts were subjected to digestion with exoglycosidase arrays, giving sequence information on the analytes and clearly identifying a structure containing a β-1 ,3-linked galactose on one of the branches, which is undigestable with the used β-1 ,4-galactosidase. Therefore, the peak corresponding to this structure is resistant to all further exoglycosidases and progressively lags behind the major peak in the electropherograms. The peaks indicated with an arrow on the figure are two components present in the internal standard mixture Genescan 500™ consisting of rhodamine-labelled oligonucleotides. From our data, we can determine the abundance ratio between these isomers.

Figure 4:

This figure illustrates the described principles in the analysis of the N-glycans of human αi-acid glycoprotein. The amount of starting material is 0,5 μg, which is equivalent with 10 pmole of said glycoprotein. The protein was deglycosylated after binding to the PVDF membrane in a well of a 96-well plate, as described under Materials and Methods. After APTS derivatisation and cleanup using the Sephadex G10-packed multiscreen approach, the sample was split in 6 equal parts and the rest of the analysis was performed as described under Figure 3. The peaks marked with an arrow on the figure are corresponding to analytes present in a mixture of rhodamine- labelled 6-,18-,30-,42-meric oligonucleotides. The observed profiles are entirely consistent with the N-glycan structures reported in reference 19 and our profiles give accurate quantitative information on the analytes, which was not possible using MALDI-TOF-MS. This exemplifies the complementarity of both technologies.

Figure 5: Sequencing of the N-glycans present on serum glycoproteins. GU: glucose units. Panel 1 : malto-oligosaccharide ladder. The number of glucose units of the 5-mer and the 10-mer is indicated above the figure. Panels 2-7: profiles obtained after digestion of serum glycoprotein N-glycans with different exoglycosidase mixtures. Panel 2: profile of the desialylated serum glycoprotein N-glycans. Panel 3: sialidase + α-1 , 3/4/6- fucosidase. Panel 4: sialidase + α-1 ,3/4-fucosidase. Panel 5: sialidase + galactosidase. Panel 6: sialidase + galactosidase + -1 ,3/4/6-fucosidase. Panel 7: same as Panel 6, but with extra β-N-acetylhexosaminidase digestion. Panels 8-10 provide the profiles obtained after the following digestions of the reference glycan of structure shown in Fig.8a. Panel 8: sialidase; Panel 9: sialidase + galactosidase; Panel 10: sialidase + galactosidase + α-1 ,3/4/6-fucosidase. Panel 11 shows the profile obtained from the reference glycan of structure shown in Fig.8b and Panel 12 the profile of the same glycan after galactosidase digest. In Panel 13 and 14, the profiles are shown of the reference glycan of structure shown in Fig.8c, with sialidase digestion and sialidase + galactosidase digestion, respectively.

Figure 6:

Profiling of the desialylated glycans of 14 CDGIa patients. GU: glucose units. Panel 1: malto-oligosaccharides. Panels 2-14: profiles of desialylated serum glycoprotein N- glycans of 14 CDGIa patients. The profiles have been arranged according to decreasing extent of deviation from the normal profile, which is shown in Panel 15. From these profiles, it can be concluded that there is increased fucosylation of the serum protein N-glycans in CDGIa, together with a decrease in the ratio between triantennary and biantennary glycans. Next to the panel numbers, the residual PMM activity as measured in fibroblasts and the clinical assessment of each patient is given.

Figure 7:

Analysis of the desialylated serum protein N-glycans of a number of patients with other subtypes of the CDGI group of syndromes. GU: glucose units. The abbreviation of the subtype is given in the top left corner of the panels. The enzymes which are deficient in these CDGI subtypes are for CDGIb: phosphomannose isomerase; CDGIc: α-1 ,3- glucosyltransferase; CDGId: α-1 ,3-mannosyltransferase; CDGIe: dolichol-phosphate- mannose synthase; CDGIf: the dolichol chaperone enzymatically equivalent to the enzyme deficient in Lee 35 CHO cells.

Figure 8:

B1 , B2, and B3 show structures of reference glycans. Detailed description of the invention

More recently, electrophoretic methods have become the analytical tools of choice for smaller quantities of carbohydrates. The introduction of fluorescent detection in combination with slab gel and capillary electrophoresis has enhanced the detectability of carbohydrates. For example, researchers have labelled mono- and oligosaccharides with fluorescent compounds and then electrophoretically separated and detected the labelled mono- and oligosaccharides. One of the earliest applications of fluorescently labelled sugars in slab gel electrophoresis was the reductive amination of sugars with 2-aminopyridine (AP). The AP reacts with the reducing end of mono- and polysaccharides. When the saccharide vicinal hydroxyls are complexed with borate at high pH (>10), or the amine functionalities are protonated at low pH <2.5), to form a charged compound, the labelled saccharide can be electrophoretically separated and detected with a fluorescent detector. Another approach to the separation and detection of carbohydrates involves the reductive amination of monosaccharides and oligosaccharides with 8-aminonaphthalene-1 ,3,6-trisulfonic acid (ANTS) followed by their electrophoretic separation and detection utilizing a CCD fluorescence imaging device. ANTS labelled oligosaccharides have been separated and detected also using capillary electrophoresis techniques. Other reagents which fluoresce and react with the reducing ends of saccharides or aminated saccharides include 5-aminonaphthalene-2- sulfonate (ANA) 3-(4-carboxybenzoyl)-2-quinolinecarboxaldehyde (CBQCA), 4- aminobenzonitrile (ABN). A widespread apparatus for the high resolution electrophoretic separation and fluorometric detection of biomolecules is the Applied Biosystems series 377 DNA-sequencer. These instruments use 12, 36 or 48 cm polyacrylamide-based gels as the separation matrix and contain an Ar-laser to excite the fluorescence of the analytes. All analytes have to pass through the same length of separation matrix, which helps to maintain resolution over a very broad range of electrophoretic mobilities. In the past, the ABI equipment has been used for the analysis of starch polymers¹2₎ where sensitivity of detection is not an issue because of the copious amounts of the analyte available. The presence of an Ar laser in the ABI sequencers and the electrophoretic separation mechanism somewhat limit the choice of the fluorescent tag used to derivatise the glycans. The fluorescence of this tag must be excitable with 488 nm light and it must contain several negative charges to obtain sufficient electrophoretic mobility of neutral glycans, which constitute an important subclass of N-glycans. In capillary electrophoresis and in the starch analysis studies using the ABI systems, 8-amino-1 ,3,6-pyrenetrisulfonic acid¹3 (APTS) has been used as it contains three sulfonic acid groups and has a λm_ax for fluorescence excitation of 434 nm after conjugation to a glycan (424nm unconjugated), with a large tail of the absorption peak extending to 488 nm. Moreover, it emits fluorescence with maximum intensity at 520 nm, close to the emission maximum of the green dyes used in DNA- sequencing. This obviates the need for the creation of a new fluorescence-overlap correction matrix for the sequencer software. An important asset of this APTS label is that sufficiently pure reagent is available from different manufacturers (Beckman Inc., Fullerton, CA, USA, Molecular Probes Inc., Eugene, CA, USA). The present invention essentially overcomes in an efficient way the problems encountered in downscaling (miniaturizing) the analysis of underivatized carbohydrates and glycoconjugates to the subpicomole level. These problems typically involve sample loss due to adsorption on container or resin walls and contamination of the sample by components present in potentially all of the reagents and vessels used in the process. For example, if a certain volume of gel filtration resin has a capacity for binding 1 pmole of APTS-labeled carbohydrates, and one applies to this resin 1 nmole of APTS-labelled carbohydrates, only 0.1% of the sample will be lost, not significantly influencing the result of the analysis. However, if only 1 pmole of APTS-labeled carbohydrates are applied, essentially all of the analytes may be bound, totally precluding the analysis of this sample. In the same way, all resins contain a certain amount of contaminating chemicals, of often unknown nature, which may be recovered in the final sample. Again, if a certain volume of a gel filtration resin contaminates the sample with 100 fmole of a fluorescent component, it will go essentially unnoticed when one applies and recovers 1 nmole of the derivatized glycans, but will seriously disturb the analysis of a 1 pmole carbohydrate-containing sample. From these arguments, it will be clear to those skilled in the art that the present invention unexpectedly overcomes the difficulties encountered in downscaling the analysis of underivatized carbohydrates, from sample quantities for which analytical methods exist in the art, to samples in which one or more of the carbohydrates is present in less than 1 pmole amount.

In a first embodiment the invention provides an electrophoretic method to analyse underivatized carbohydrates present in less than 1 picomole amount in a sample. The present electrophoretic method is advantageous in that it is possible to start with less than 1 picomole amount of underivatized carbohydrates, such as for example glycosylated proteins, present in a sample, thus effectively increasing the sensitivity of the overall analytical process one to two orders of magnitude as compared to the situation in the prior art.

In another embodiment the electrophoretic method comprises: (1) derivatizing the carbohydrates with a derivatizing agent, (2) removing the portion of the derivatizing agent which did not derivatize with said carbohydrate, (3) separating the derivatized carbohydrates and (4) the detection of the derivatized carbohydrates. According to the invention, the carbohydrates can be enzymatically and/or chemically released from a glycoconjugate using a miniaturized procedure. After release, the carbohydrates are derivatized with APTS. An essential feature of this invention is that the APTS label can be efficiently separated from the labelled carbohydrates in order not to interfere with the background of the detection due to overloading of the analytical separation system. In other words, in methods described in the art, it is invariably necessary to use a large excess of fluorescent tag, APTS, in order for quantitative derivatisation to occur. This implies that this excessive amount of tag must be removed if trace amounts of derivatised glycans are to be detected. This problem is in current methods, available in the art, circumvented by derivatising large amounts of glycans and diluting the derivatisation mixture prior to analysis. This may result in high sensitivity, but it is not useful where the amounts of starting material are extremely low. Oligosaccharides can also be efficiently detected by MALDI-TOF-MS and no derivatisation is needed but here highly efficient desalting procedures are required and this also gets more complicated in trace analysis. Another problem relating to MALDI-TOF-MS of glycans is that the use of the results for quantitation of the compounds is still a matter of debate and that isobaric (regio-and sterioisomeric) structures, which are very common in carbohydrate structures, cannot be resolved.

Thus, the present invention provides an improved analytical method for analyzing sugars and carbohydrates including mono-, oligo- and polysaccharides as well as sugar or carbohydrate containing compounds. A 'glycoconjugate' means any compound containing a carbohydrate moiety. By the term 'carbohydrate' it is meant a collection of molecules which can be arbitrarily subdived into four major groups- monosaccharides, disaccharides, oligosaccharides, and polysaccharides and derivatives of the members of these four groups. Three of the most important monosaccharides or simple sugars, mannose, glucose and galactose, have the same molecular formula, (C6H12O6), but, because their atoms have different structural arrangements, the sugars have different characteristics; i.e., they are isomers. Slight changes in structural arrangements are detectable by living things and influence the biological significance of isomeric compounds. Two molecules of a monosaccharide or derivatives thereof that are linked to each other form a disaccharide, or double sugar. An example of a disaccharide is sucrose. Oligosaccharides consist of three to six monosaccharide units or derivatives thereof. Polysaccharides (literally this term means many sugars) consist of more than six units of monosaccharides or derivatives thereof and represent most of the structural and energy-reserve carbohydrates found in nature. Large molecules that may consist of as many as 10,000 monosaccharide units or derivatives thereof linked together, polysaccharides vary considerably in size, in structural complexity, and in sugar content; several hundred distinct types have thus far been identified. Cellulose, the principal structural component of plants, is a complex polysaccharide comprising many glucose units linked together; it is the most common polysaccharide. The starch found in plants and the glycogen found in animals also are complex glucose polysaccharides. The words "glycan" and "carbohydrate" are interchangeably.

Especially, with the analytical method of the present invention the carbohydrates can belong to any class of protein or lipid linked carbohydrates comprising asparagine- linked glycans, Ser/Thr-linked glycans, glycosaminoglycans or proteoglycan derived sugars, glycolipid-derived glycans and GPI-anchor derived carbohydrate species. The processes of the present invention provide enhanced detectability of femtomole amounts of glycans. Furthermore, during electrophoretic separation the standard buffer used for DNA-sequencing gels was used throughout. The borate in this buffer forms complexes of different stability with different carbohydrate isomers increases the chance that isomers can be electrophoretically resolved (as exemplified in Figure 3). When referring to 'to analyse underivatized carbohydrates', carbohydrates can be bound or not to another molecule. In the present invention 1 picomole or less of underivatived carbohydrates can be analysed. The latter can mean that a sample comprising carbohydrates contains 1 picomole or less total carbohydrates or a specific carbohydrate of interest is present in less than 1 picomole in said sample. The latter means that in a non-limiting example if total blood serum is analysed for diagnosis (which normally comprises more than 1 picomole amounts of underivatized carbohydrates) of a specific disorder that a disease specific carbohydrate structure is often present in only 1 picomole or less of said carbohydrate and it can still be analysed. Preferably said carbohydrate analysis can be efficiently carried out on 1 , 0.5, 0.1 , 0.01 or 0.001 picomole amounts of carbohydrates present in a sample. However, it should also be clear that higher amounts of free carbohydrates can be analysed with this method such as 2, 5, 10, 50, 100, 500, 1000, 10.000 picomole or even higher amounts.

If further characterization is needed, such as sequencing, less than 1 picomole of starting glycoconjugate is needed to obtain results for each extra exoglycosidase array. The term 'sample' comprises, but is not limited to, a specific sugar or a mixture of sugars in solution, a blood sample, a cell extract, urine, sperm, micro-dissected cells, cerebrospinal fluid, industrial water and sewage.

In the present invention the efficient separation analysis of the derivatized carbohydrates is carried out by an electrophoretic process and the standard DNA- sequencing equipment can be used. The high sensitivity of the overall analytical process, is obtained using known in the art deglycosylation procedures, a derivatisation step with APTS, a highly efficient and reproducible novel post- derivatisation cleanup step to remove the APTS involving Sephadex G10. The novel post-derivatisation step described here in the invention is surprisingly efficient considering 1) the enormous excess of free label that is efficiently removed and 2) the small difference in molecular size between the free label and the labelled analytes. To those skilled in the art, it will be clear that this unexpected finding could be carried out by for example spin columns packed with other suitable size selective media known in the art as a gel filtration or gel permeation resin, or using a size-selective membrane. It should be said that removal of APTS by means of gel filtration has been described in the art but samples of 1 nanomole or more were applied to the resin. In a preferred embodiment this novel method can be made medium-throughput by performing all sample preparation steps (enzymatic deglycosylation with PNGaseF, desalting, derivatisation with APTS and post-derivatisation cleanup) in 96-well based plates. This integrated sample preparation scheme is compatible with capillary electrophoresis platforms already in use. With the described technology high- performance glycosylation analysis is brought within reach of each life sciences lab. The present invention provides methods for the electrophoretic separation and detection of saccharides including simple sugars, oligosaccharides, carbohydrates which or free or bound to other molecules (glycoconjugate). A wide range of analytical applications can take benefit from the present invention, including carbohydrate and glycoprotein sequencing, industrial sugar and carbohydrate analytical procedures, drug analysis, drug discovery, and diagnostic procedures. In particular, those skilled in the art will appreciate that the ability to readily analyse less than 1 picomole amount of underivatized carbohydrates by a method which gives complementary information to mass spectrometry is very desirable in the analysis of carbohydrate and glycoconjugate sequencing reaction products and the diagnostic analysis of biological samples.

Although the present invention provides a method for the analysis of glycoproteins in combination with a DNA-sequencer it is clear for the person skilled in the art that this method can be applied in connection with capillary electrophoresis systems adaptable to a laser induced fluorescence detector. Such systems include the P/ACE series Capillary Electrophoresis Systems (Beckman Instruments, Inc., Fullerton, Calif.). However, the processes described herein can be applied with any electrophoresis system which is adaptable with a laser induced fluorescence detector. It is also clear that an analytical system that has the same overall sensitivity as MALDI- TOF-MS and gives complementary information such as resolution of isobars and accurate quantitation is of utmost importance in the field of glycan analysis. The term "derivatized carbohydrate compounds" means carbohydrates which have been labelled with a fluorescing compound. In order for analytes to migrate under electrophoretic conditions they must carry a charge and since many carbohydrates are not charged, the fluorescing compounds are preferably charged. Fluorescing compounds such as 9-aminopyrene-1 ,4,6-trisulfonic acid (APTS) and 8- aminonaphthalene-1 ,3,6-trisulfonate (ANTS) are particularly suitable in the practice of the present invention. Regarding the detection of the derivatized carbohydrates, any detection method known in the art can be applied, preferably the detection is carried out with a laser such as a diode laser, a He/Cd laser or an argon-ion laser.

In another embodiment the invention provides a device for the detection of less than 1 picomole of underivatized carbohydrates comprising: (1) a derivatization chamber, (2) a clean-up apparatus wherein said clean-up occurs via gel filtration and/or gel permeation resin, in which the fluid flow is driven by centrifugational force or vacuum, which remove the excess fluorescent and/or spectrometric label, and (3) an electrophoretic apparatus such as a capillary electrophoresis and/or DNA-sequencing equipment for the separation and detection of derivatized carbohydrates. In another embodiment the carbohydrate analysis method can be supplemented pre- electrophoretically with an internal standard mixture that is labelled with a chromophore or fluorophore that is different from the label attached to the carbohydrate analytes. The internal standard allows for accurate and reproducible determination of the electrophoretic mobilities of the carbohydrate analytes by referencing these mobilities to the mobilities of the components in the internal standard mixture. In the present invention, for example, a rhodamine-labelled oligonucleotide standard Genescan™ 500 (Applied Biosystems, Foster City, CA, USA) or a mixture of rhodamine-labelled 6-,18-,30-,and 42-meric oligonucleotides may be added to the derivatised glycans before profiling.

In another embodiment the analysis method of this invention can be used to distinguish recombinant glycoproteins form their endogenous counterparts on the basis of differences in their protein-linked glycosylation as is exemplified herein. In another embodiment the method of the present invention can be used for diagnosing diseases where at least one modification of one or more carbohydrates is implicated. In other words the analysis method of this invention can be used for sensitive diagnostic purposes for disorders such as carbohydrate storage disorders, carbohydrate-deficient glycoprotein syndromes, cancer, mood disorders, disorders in the biosynthesis of protein or lipid-linked glycans and more generally in any field involving changes in carbohydrate profiles.

In another embodiment the method of the invention can be used for the identification of carbohydrate structures of recombinant glycoproteins in biological fluids. In a specific embodiment the method can be used to differentiate between endogenously made erythropoetin and recombinant erythropoetin by analysing the carbohydrate structures of erythropoetin.

In yet another embodiment the analysis method of this invention can be used for the identification and/or structural characterization of carbohydrates which are bound to other biomolecules. By 'biomolecules' it is meant molecules comprising nucleic acids, proteins, other carbohydrates and lectins. In yet another embodiment this invention can be used for the analysis of carbohydrates derived from organisms comprising prions, viruses, bacteria, fungi, mycoplasma and parasites. By parasites it is meant organisms such as for example protozoa and worms. Finally, in another embodiment the analysis method of this invention can be used for obtaining the information on the structure of the carbohydrates (this generally refers to sequencing of carbohydrate structures) by combining the current analysis method with a procedure known to those skilled in the art as chemical and exoglycosidase sequencing and modifications thereof.

The examples which follow are offered as descriptive of certain embodiments. As such they are exemplary only and are not limiting in their nature.

Examples 1. Experimental design of the carbohydrate analysis

Experiments showed us that overloading of the gel was observed if more than 100 picomol of the free label of APTS was present in the sample of derivatized carbohydrates. The minimal concentration of APTS necessary to obtain quantitative derivatisation in a reasonable time span (overnight incubation at 37°C) is about 10 mM¹⁴. The derivatisation reactions in this study were miniaturized to 1 μl volumes (performed in 250 μl ultraclean PCR tubes), which means that 10 nanomol of the label is present. From these considerations, it is evident that a cleanup methodology would be highly desirable that can remove more than 95% of the label in order to be able to load a significant fraction of the labelled carbohydrates (± 20%) on a sequencing gel lane. For this purpose, we tried several approaches (paper chromatography, thin layer chromatography, reaction with partially oxidized Sephadex G75 beads¹5), but none of those worked. Finally we turned to an approach using Sephadex G10 packed spin columns. We initially tested the cleanup efficiency of 1 ,5 cm beds of Sephadex G10, packed in microspin columns. Surprisingly, over 95% of the free APTS label could easily and reproducibly be removed, with a recovery of >70% of labelled ¹⁴C-lactose (disaccharide) and ¹ C-sialyllactose (trisaccharide), the neutral and sialylated test compounds used. Subsequently, we adapted this cleanup procedure to a 96-well format by filling the wells of a 96-well Durapore-lined Multiscreen plate (Millipore, Bedford, CA, USA) with the Sephadex G10 resin. The eluate of these plates is collected in another 96-well plate with tapered wells and evaporated to dryness in a vacuum centrifuge equipped for 96-well plates. The labelled glycans are then reconstituted in water. To increase the reproducibility and accuracy of the glycan profiling, we add the rhodamine-labelled oligonucleotide standard Genescan™ 500 (Applied Biosystems, Foster City, CA, USA) to each sample. This results in the detection of two major peaks in the range of electrophoretic mobilities relevant here. By reserving one lane of each gel for an APTS-derivatised malto-oligosaccharide ladder, also containing the Genescan™ 500 standard, the electrophoretic mobility of each glycan can be very reproducibly expressed in glucose units (GU's). This kind of internal referencing of carbohydrate analysis profiles has so far only been described in High Performance Anion Exchange Chromatography, where both a Pulsed Amperometric Detector (detection of the malto-oligosaccharides) and a fluorescence detector (detection of labelled analytes) were necessary to obtain this result¹6. Here, the four-colour fluorescence capabilities of the DNA-sequencer obviate this need. The internal standardisation principle introduced in this invention is implementable on the capillary electrophoresis systems that are equipped with multi-colour fluorescent detection systems. The acrylamide percentage of the gel used here (12%) and the other electrophoresis conditions were optimized for maximum resolution of a malto- oligosaccharide reference mixture with degrees of polymerization of 4-25. This is the size range that is most relevant for N-glycan mixtures derived from mammalian and plant tissues. The standard buffer used for DNA-sequencing gels was used throughout (see Experimental protocol). The borate in this buffer forms complexes of different stability with different carbohydrate isomers, thus increasing the chance of electrophoretically resolving these isomers. It should be straightforward to optimize the electrophoresis parameters for other classes of protein-linked glycans, if necessary. As shown in Fig. 2, the detector-response curve is linear over more than three orders of magnitude and 1 fmol of labelled chitotetraose (test compound) can be detected with a signal to noise ratio of >3. The 96-well based cleanup-procedure of APTS-derivatized glycans described here is also applicable for capillary electrophoresis of these compounds^ and thus allows the utilization of the full potential sensitivity of this methodology (for example, as commercialized by Beckman Inc., Fullerton, CA, USA). MALDI-TOF-MS of underivatized N-linked glycans is a well established technique⁴.¹^-¹9 _at about the same level of sensitivity as the DNA- sequencer-assisted methodology descirbed here. The two techniques can give complementary information on the analytes. The sequencer technique often resolves the isobaric glycan isomers and provides reliable quantitation of the observed species, whereas MALDI-TOF-MS gives the exact mass of the glycans. Therefore, we adapted the AG-50-WX8 desalting step described by Papac et al ⁴ to a 96-well format (see Experimental Protocol), completing our high-throughput sample preparation scheme. Combination of the described approaches with exoglycosidase digestions can give structural information on the glycans under study at the low femtomolar level. In Fig 3a, this is exemplified for a mixture of trisialylated triantennary complex type N-linked oligosaccharides. In Fig. 3b, we show the N-glycan profiling of glycans derived from 500 ng αi-acid glycoprotein (the equivalent of about 10 pmole protein). This amount of starting material is sufficient to obtain both the native profile and the results of five exoglycosidase arrays.

2. Sequencing of femtomole amounts of carbohydrates derived from low picomole amounts of glycoproteins

The present methodology can be combined with carbohydrate release procedures described in the literature ⁴ allowing the reproducible release of the N-linked glycans from low picomole amounts of glycoproteins, and with sequencing procedures of the glycans under study with arrays of exoglycosidases, thus generating a wealth of structural information on the glycans from an amount of starting material, the analysis of which was previously only feasible with MALDI-TOF mass spectrometry. However, MALDI-TOF MS does not resolve isobaric structures, which are very prevalent in the glycan field, and is not absolutely accurate in the quantification of the analytes. These problems are solved in the present method, making the information obtained complementary with the data from MALDI-TOF MS. The high sensitivity of the methodology gives access to the study of glycosylation-related research topics that were previously not feasible. For example, it is now feasible to study the glycosylation potential of rare cell populations in the bloodstream and in the nervous system and the changes in this potential during development or during disease progress.

3. Structural characterization of the sugar epitopes recognized by lectins In the field of lectinology, it is not possible to reliably predict the carbohydrate specificity from the gene sequence. With the advent of whole-genome sequencing, a lot of lectin-domain containing sequences are being discovered, some of them with crucial functions such as animal development, inflammation and host-pathogen recognition. In order to design specific inhibitors for these crucial interactions, it is vital to be able to obtain structural information on carbohydrate epitopes binding to these lectin domains. A method of the highest sensitivity and resolution is very important in this respect, as most of these proteins will only be available in limited amounts. One can envisage procedures in which binding carbohydrates are isolated from a complex pool of labelled glycans by affinity procedures with the lectin, followed by structural studies of the binding carbohydrates with exoglycosidase sequencing. This structural information can then be used in drug discovery processes aimed at inhibiting the lectin-carbohydrate interaction under study.

4. Detection of recombinant glycoproteins after administration in vivo

The method discussed here can be used for the detection of recombinant glycoproteins in vivo if the glycosylation of the recombinant glycoprotein differs from the glycosylation of its endogenous equivalent. For example, this is the case with recombinant erythropoietin, IFN-gamma and tissue plasminogen activator which are derived from CHO cells and their native counterpart. As the profiles obtained from endogenous and recombinant glycoproteins are generally different, the latter can be reliably detected with our invention. In the case of EPO, this can allow for a more reliable test to be developed to detect the inappropriate use of recombinant EPO, for example in sports competitions.

For example, it is well known that the carbohydrates structures of recombinant glycoproteins from mammalian non-human and non-chicken cell lines or organisms, such as recombinant human EPO (rhuEPO) derived from Chinese Hamster Ovary (CHO) or Baby Hamster Kidney (BHK), contain varying amounts of 5-N- glycolylneuraminic acid. In the case of recombinant human EPO, this amount is 1-2% of total sialic acids present on the N-glycans of the protein (HOKKE CH, BERGWERFF AA, VANDEDEM GWK, KAMERLING JP, VLIEGENTHART JF.EUROPEAN JOURNAL OF BIOCHEMISTRY 228: (3) 981-1008 MAR 15 1995). This particular glycosylation feature is undetectable on human blood plasma proteins (Muchmore et al. Am. J. Phys. Anthropol. 107, 187-198, 1998).

One can thus envisage the detection of recombinant glycoproteins, such as rhuEPO, in a human by collecting a sample, such as blood or urine, from the person to be controlled. A glycosylation analysis is then performed on total plasma or serum proteins or preparations thereof.

Alternatively, this glycosylation analysis can be performed after a pre-purification of the analytes of interest. For example, erythropoietin present in the sample is enriched using affinity techniques involving anti-erythropoietin antibodies or recombinant erythropoeitin receptors (Zhan et al., Protein Engineering, 12: (6), 503-513, 1999) or preparations thereof, after which a glycosylation analysis is performed. Another embodiment can consist of isolating the protein-linked carbohydrates from all or a fraction of the proteins present in the said sample, followed by enrichment of the 5-N- glycolylneuraminic acid-containing saccharides using lectins and preparations thereof specifically recognizing said glycosylation epitope. Such lectins include those isolated from Morus alba (Ratanapo-Sunanta et al., Plant Science Shannon 139 (2), 141-148, 1998), Pila globosa (Swarnakar et al., Biochemical and Biophysical Research Communications 178 (1), 85-94, 1991), Scylla serrata (Mercy et al. Eur.J.Biochem 215 (3) 697-704, 1993) and Anadara granosa (Dam et al., Biochemical and Biophysical Research Communications 196 (1), 422-429, 1993).

In the particular example of recombinant human EPO (rhuEPO), this glycoprotein is injected intravenously or subcutaneously in varying doses, normally several times per week for an extended period. A typical treatment may consist of subcutaneous injection of 100 IU rhuEPO per kg body weight three times per week. As 1 IU of rhuEPO is equivalent with about 10 ng of the glycoprotein, the dose corresponds to about 70 microgram. Calculating a molecular weight of about 35 kDa, this dose is equivalent to about 2 nmol of rhuEPO. Further calculating a blood volume of 5 liter one obtains a value of 0,4 pmol per ml blood. This does not take into account the distribution of the injected glycoprotein between the blood compartment and the rest of the body. Assuming that 25 % of the injected EPO distributes to the blood compartment, we get a figure of about 0,1 pmol per ml. Also, it has been determined that the mean elimination halflife of rhuEPO in athletes is about 42 h (Bressolle F, Audran M, Gareau R, Baynes RD, Guidicelli C, Gomeni R, CLINICAL DRUG INVESTIGATION 14: (3) 233-242 SEP 1997). Assuming that sampling takes place within 42 h after injection, the amount of rhuEPO per ml blood is calculated to be about 50 fmol. In a blood sample of 10 ml, one expects 500 fmol of the drug. Each rhuEPO molecule contains 3 N-glycans that can be removed from the protein using chemical or enzymatic reactions. One can thus expect a yield of 1 ,5 pmol of total N-glycans originating from the rhuEPO. As mentioned, about 1-2 % of these glycans bear at least one 5-N-glycolylneuraminic acid modification. So, one expects about 20 fmol of these specific, non-derivatised glycans to be present in the sample, an amount of carbohydrates only possible to analyse with the present invention, even when assuming the presence of several isomeric structures containing this 5-N- glycolylneuraminic acid.

Other applications are expected in the field of the study of pharmacokinetic behaviour of recombinant protein therapeutics with an endogenous equivalent or in the definition of glycoforms related to new or unwanted functions of the glycoprotein. In most of these cases, the glycosylation profile is the only difference between recombinant and endogenous protein.

4.1 Analysis of subpicomole amounts of recombinant human erythropoietin (EPO) produced in CHO cells

Recombinant human EPO (rhuEPO) was purchased from Roche (Cat. No. 1120166). 1 unit of this preparation equals about 10-15 ng protein. A dilution series of 5,2,1 ,0.5 and 0.1 units EPO was analysed. The entire procedure was as described in the current patent application except that 50 mM APTS was used in the derivatization procedure and that the cleanup over the microtiter-Sephadex G10 bed was carried out two consecutive times, with volume reduction of the first eluate using vacuum evaporation. After cleanup, half of the labelled glycans were digested with Arthrobacter ureafaciens sialidase at 2 U/ml and alpha-fucosidase from bovine kidney at 0.5 U/ml. This procedure results in a homogenisation of the N-glycan profile and is especially useful in measuring the amount of glycans with N-acetyllactosamino repeats, a particular hallmark of recombinant EPO. In rhuEPO, about 30% of the tetraantennary glycans have one N-acetyllactosamino repeat, whereas this is only 10% for endogenously synthesized human EPO. The tetraantennary glycan is the major peak in the profiles, whereas the N-acetyllactosamino-derivative is the second tallest peak. The ratio between the two peaks is easily quantifiable with an amount of starting material down to 0.5 units (which corresponds with only 100 femtomole EPO), making this assay useful for purposes where distinction between endogenous and recombinant EPO is necessary, such as in detection of abuse of this protein in sports competitions or in monitoring the EPO source in treatment of patients with chronic kidney insufficiency.

5. Detection of disease-related glycosylation alterations for diagnostic purposes Many disease states are hallmarked by alterations in the glycosylation of specific proteins, lipids or tissues or by the accumulation of specific products of carbohydrate nature. Well-documented examples are the increased exposure of terminal beta-1 ,4- galactose residues in rheumatoid arthritis IgG, the increased beta-1 ,6-branching of N- linked carbohydrates in numerous tumor tissues and the accumulation of high- mannose oligosaccharides in the serum and urine of alpha-mannosidosis patients. Using the technology described here, only very small specimens derived from the patient are necessary to provide detailed information on the carbohydrates involved. Moreover, we disclose methods for high throughput sample preparation of these carbohydrates, which are very important in the clinical setting.

5.1 Increased fucosylation and reduced branching of serum glycoprotein N-glycans in all known subtypes of Congenital Disorder of Glycosylation I

Congenital Disorders of Glycosylation are a group of inherited disorders characterized by alterations in protein glycosylation. CDG I is hallmarked by the absence of one or more N-glycan chains on serum glycoproteins. Several subtypes have been described, of which the most common is CDG la, caused by mutations in the PMM2 gene, which result in deficiency in phosphomannomutase activity. This enzyme is essential for the synthesis of GDP-mannose, a substrate that is indispensable in the biosynthesis of N- linked glycans. CDG I is commonly diagnosed using isoelectric focusing of serum proteins, followed by immunodetection of transferrine isoforms. This technique basically detects different sialoforms of transferrine and has been extensively described. However, it does not give detailed structural information on the N-glycans present. In this example, we perform a quantitative analysis of the N-glycans present on the total mixture of serum proteins of CDG I patients as compared to those of persons with normal transferrine patterns. We show that hyperfucosylation of the core- and branch type occurs in CDG I and that tri-antennary and tetra-antennary glycans have a lower relative abundance, as compared to biantennary glycans. Moreover, for CDGIa the extent of these changes correlates relatively well with the clinical presentation of the patients and may be used as a prediction marker for the clinical outcome of the disease, which has been difficult using other techniques, (a) analysis of N-glycans present on serum glycoproteins of normal human serum. The N-glycans of a representative normal serum was profiled and a surprisingly simple fingerprint arose after desialylation of the glycans using the broad-specificity

^• Arthrobacter ureafaciens sialidase (Fig.1 Panel 2). For the purpose of this paper, only four major peaks are of paramount importance and we limited the structural information to these four peaks. From comparison with oligosaccharide standards and from the analysis of the profiles after digestion with different exoglycosidase arrays, we could conclude that the main peaks represented biantennary and triantennary glycans, either unfucosylated or substituted with one fucose residue. Peak 1: this is the most abundant N-glycan on total serum glycoproteins. After sialidase digest (Fig.1 , panel 2), it's size is estimated to be 9 monosaccharide units from comparison with the malto- oligosaccharide reference ladder. After sialidase/β-1 ,4-galactosidase digest (Fig. 1 , panel 5), this glycan loses two galactose residues and it further loses two GlcNAc residues upon digestion with β-N-acetylhexosaminidase (Fig.1 , panel 7). The residual glycan migrates at the position of the Man₃GlcNAc₂ core N-glycan structure. We conclude that the structure of glycan nr.1 is biantennary, bi-β-1 ,4-galactosylated. This conclusion is corroborated by its exact comigration with a reference glycan of this structure, both undigested (compare panel 2 and panel 11) and digested with β-1 ,4- galactosidase (compare panel 5 and panel 12). Peak 2: this glycan is about 1 monosaccharide unit larger than the glycan corresponding to Peak 1. Upon bovine kidney fucosidase digest (Panel 3), this glycan is converted to glycan nr.1 , as no new peak appears in the post-fucosidase profile that could account for a glycan with the large abundance or Peak 2. After removal of the two β-1 ,4-linked galactose residues, the same conclusion can be drawn from comparison of Panel 5 and Panel 6, where the peak marked as 2' is converted in the peak marked with 1'. Peak 2 is resistant to digestion with the almond meal α-1 ,3/4-fucosidase, indicating that the fucose residue is of the α-1 ,6 core type. Thus, Peak 2 represents the biantennary, bi-β-1 ,4- galactosylated core α-1 ,6-fucosylated glycan. This is corroborated by the exact comigration of a reference glycan of this structure with peak 2, both undigested and after β-1 ,4-galactosidase and β-1 ,4-galactosidase/bovine kidney fucosidase double digest (compare panel 2 and panel 8; panel 5 and panel 9; panel 6 and panel 10, respectively). Peak 3: the glycan corresponding to peak 3 is about 2 monosaccharide units taller than glycan 1 , is not digestable by bovine kidney fucosidase and comigrates with a triantennary fully β-1 ,4-galactosylated reference glycan. Moreover, β-1 ,4- galactosidase removes 3 galactose residues from the glycan (shift of 3 glucose units between panel 2 and panel 6), after which the glycan is 1 monosaccharide unit taller than the remnant of glycan 1 , in accordance with the one extra GlcNAc residue that is expected for a triantennary glycan when compared to a biantennary structure. Peak 4: this glycan is one glucose unit taller than the triantennary unfucosylated glycan of peak 3 and is sensitive to both bovine kidney (panel 3) and almond meal fucosidase (panel 4), after which digests the glycan is converted to peak 3. Thus, the fucose residue present on this glycan is α-1 , 3/4 linked. We conclude that the glycan of peak 4 is a branch-fucosylated derivative of glycan 3. The exact position of the branch fucose residue cannot be determined using enzyme digests. (b) Analysis of CDGIa patient sera. 14 genetically characterized CDGIa patients were selected for analysis. The patient data (mutation type and residual PMM activity) are shown in figure 2. After desialylation of the glycans, the profiles shown in Fig. 2, panel 2-14 were obtained. For comparison, panel 15 is the profile obtained from a normal serum. Qualitatively, the same profile is obtained as for the normal serum, but important quantitative differences were observed. Four main changes are apparent : first, the peak indicated as 0 in Fig.2 panel 1 is significantly upregulated. This glycan was found to have the core-fucosylated agalacto biantennary structure (digests not shown) and indicates a somewhat less efficient galactosylation by the liver in CDGIa. Second, the extent of core fucosylation on the biantennary glycans is increased (P=0,02), as can be seen from the ratio of peaks 0 and 2 to peak 1. Third, the amount of triantennary glycans (peak 3 and 4) was decreased (P= 0,0005) with respect to the biantennary glycans (peaks 0,1 and 2). Fourth, although the total population of triantennary glycans was less abundant than in normal sera, a larger percentage (P=0,0002) of these triantennary glycans carried the branch fucose residue. (c) analysis of the other known CDGI subtypes. Congenital Disorders of Glycosylation type I are defined as the group of disorders with malfunctioning of the ER-localized steps in the glycan biosynthesis pathway. All of the subtypes known so far affect the biosynthesis of the Glc₃MangGlcNAc₂-dolichol N-glycan precursor, resulting in a less efficient occupation of the N-glycosylation sites on nascent proteins in the ER. So, what the Golgi N-glycosylation biosynthesis pathway 'sees' in this group of diseases would be similar in all of these syndromes and therefore, we assumed that similar quantitative changes as observed in CDGIa could occur in the other CDGI subtypes, as all of these subtypes result in a reduction in the number of occupied N-glycosylation sites. Therefore, we analysed the serum glycoprotein N-glycans of one patient with CDGIb (phosphomannose isomerase deficiency), two with CDGIc, two with CDGId, three with CDGIe and two with CDGIf. For most of these disorders, the limited number of known patient precludes sound statistical analysis. However, it is apparent from the profiles in Fig. 3 that the same quantitative changes as in CDGIa occur in these profiles. Statistical comparison of the pooled group of profiles of the non-CDGIa disorders (n=10) with a group of normal profiles (n=8) revealed increased core fucosylation of the biantennary glycans (P=0,005), a lower ratio of triantennary glycans to biantennary glycans (P=0, 000001) and an increased ratio of branch-fucosylated triantennary glycans to unfucosylated triantennary glycans (P=0, 00002). (d) correlation of the severity of the 'glyco' phenotype with the clinical severity of CDGIa symptoms. In Figure 2, the profiles obtained from CDGIa patients have been sorted in 'decreasing order of phenotype severity', compared to normal sera. It is evident that there is a gradient of phenotypes in the tested patient population and therefore, it is interesting to see if this gradient matches some other characteristic biochemical parameter of the disease or a clinically observable phenotype. To be able to create an objective score of the extent of deviation of the N-glycan profiles from the average normal profile, we determined the one-tailed 99% confidence intervals for the three most obvious N-glycan structural changes : 1) increased core-fucosylation of biantennary glycans, 2) decreased ratio of triantennary to biantennary glycans and 3) increased ratio of branch-fucosylated to unfucosylated triantennary glycans. For each of these three parameters a patient profile got a score of 1 if the parameter value exceeded the 99% confidence interval. Subsequently, the scores for the three parameters were added. So, a patient can get a score ranging from 0 to 3 with increasing deviation of the normal profile. These scores are shown in Fig. 2. An obvious biochemical parameter that could correlate with the 'glyco'-phenotype is the residual PMM enzymatic activity. The residual PMM activity as measured in patient skin fibroblasts is indicated in Fig.2. It is apparent that there is a tendency for the PMM activity to be low when the serum N-glycan profile differs severely from the normal one, but the correlation is by no means perfect. However, the standard deviation for the PMM activity measurements is very large, which may hamper correlation analysis. Subsequently, we compared the obtained profile scores with the clinical presentation of the patients. The symptoms of CDGIa can vary widely from patient to patient from a severely debilitating disease to a relatively mild mental retardation. It has been observed before that the severity of the disease does not correlate well with residual PMM activity. The clinical presentation of the CDGIa patients studied here is indicated in Fig.2 in terms of 'severe', 'moderate to severe', 'mild' and 'very mild'. The patients were classified in these categories by experienced clinicians of the European CDGI national reference centers. For two non-European patients, relatives of each other and both with a low score for their serum glycoprofiles, no clinical information could be obtained. From comparison of the profile scores with the clinical evaluation, it can be concluded that mild cases of the disease also have a low profile score. N-glycan analytical methods: (1) Sample preparation procedure: 5 μl of sera of genetically confirmed CDG la (n=14), CDG lb (n=1), CDG lc (n=2), CDG Id (n=2), CDGIe (n=3), CDG If (n=2) patients and individuals with normal serum transferrine IEF pattern (n=8) were incubated with 50 μl of RCM buffer (8M urea, 360 mM Tris, pH 8.6, 3.2 mM EDTA) at 50 °C for 1h to denature the serum proteins. Subsequently, these mixtures were loaded in the wells of a Multiscreen-IP plate (Millipore, Bedford, CA, USA), prepared as described herein. Reduction, iodoalkylation and deglycosylation steps were performed according to methods known in the art. (2) APTS derivatization reaction and cleanup: N-glycan derivatisation with 8-amino-1 ,3,6-pyrenetrisulfonic acid and removal of excess free label were as described recently. Briefly, the deglycosylation mixture was evaporated to dryness and a 1 μl 1 :1 mixture of 20 mM APTS (Molecular Probes, Eugene, CA, USA) in 1.2 M citric acid and 1 M NaCNBH₃ in DMSO was added. The derivatisation was allowed for 18h at 37°C. After this, the reaction was quenched by the addition of 10 μl of distilled (Dl) water. Excess not reacted APTS was removed using a bed of Sephadex G10 packed in a Multiscreen filterplate (Millipore, Bedford, CA, USA). After sample application, the resin beds were eluted three times by addition of 10 μl of water and a 10 second centrifugation at 750 x g in a table-top centrifuge equipped for handling 96-well plates. The eluate is evaporated to dryness. After evaporation, the derivatized glycans are reconstituted in 5 μl of Dl water. (3) Exoglycosidase digestions: 1 μl batches of the cleaned-up derivatized N-glycans were transferred to 250 μl PCR tubes or tapered-well microtiter plates for treatment with exoglycosidase arrays. In this example all digestions were done by overnight incubation at 37°C in 10 μl 20 mM sodium acetate pH 5.0. The enzymes used in this study are: Arthrobacter ureafaciens sialidase (2 U/ml, Boehringer, Mannheim, Germany); Diplococcus pneumoniae β-1 ,4-galactosidase (1 U/ml, Boehringer, Mannheim, Germany); Jack bean β-N-acetylhexosaminidase (30 U/ml, Glyko, Novato, CA, USA); Jack bean α-mannosidase (100 U/ml, Sigma Biochemicals, Bornem, Belgium); bovine epididymis α-fucosidase (Glyko, Novato, CA, USA) and almond fucosidase (Glyko, Novato, CA). After completion of the digestions, the samples were evaporated to dryness and reconstituted in 1 μl Dl water. (4) Analysis by DNA-sequencer: to each sample, 0,5 μl of the ROX-labelled Genescan™ 500 standard mixture (Perkin Elmer, Foster City, CA, USA) was added. After addition of the internal standard, 1 μl of deionized formamide was added to facilitate sample loading. All experiments were performed on an Applied Biosystems 377A DNA- sequencer (Perkin Elmer, Foster City, CA, USA), adapted for cooling as described. The 36 cm gel contained 10% of a 19:1 mixture of acrylamide.bisacrylamide (89 mM Tris, 89 mM borate, 2.2 mM EDTA). Prerunning was done at 3000 V for 1h. The electrophoresis voltage during separation was 3500 V and data were collected for 3h (separation of glycans up to 15 glucose units in size). Data analysis was performed using the Genescan 3.1 software (Applied Biosystems, Foster City, CA, USA). Using the positions of the peaks of the internal ROX-oligonucleotide standard, all lanes on the same gel were aligned with the lane containing the APTS-labelled malto- oligosaccharide standard. After this alignment, samples on different gels can be easily and reliably compared by aligning the positions of the malto-oligosaccharides present on both gels. For clarity and to allow black-and white reproduction of the figures presented in this contribution, the peaks corresponding to the ROX-labelled internal standards have been omitted after the alignment procedure.

6. Detection of changes in glycosylation upon treatment with drugs and use thereof. The present methodology can be used for the follow-up of drug-induced glycosylation changes. For example one can envisage the use of the disclosed technology to follow up the efficacy of enzyme replacement therapy or gene therapy in lysosomal storage diseases or carbohydrate deficient glycoprotein syndromes. Another example could be the analysis of the effect of potential inhibitors of glycosyltransferases or glycosidases on the glycosylation profile displayed by tumor cells, both in vitro and in vivo. For example, the exquisite sensitivity of the disclosed methodology only requires small biopsies of tumor tissue to be made.

7. Application of the invention for the profiling of glycosaminoglycans ^* Glycosaminoglycans are a class of polysaccharides with an extremely complex structure and several important biological functions, some of which are influenced by the exact patterns of sulphation and the exact sequence of the glycan chain. Their role as specific ligands for proteins has recently become apparent and this binding often modulates the function of the bound protein. Structure elucidation of the GAG fragments which bind to specific proteins can be accomplished by mass spectrometry and by chromatographic and electrophoretic methodologies. In particular, the use of specific glycosidases is paramount to the exact sequence determination of the GAG fragments. This technology has been recently developed for heparin and heparan sulphate. These specific GAGs consist of a disaccharide repeat of glucosamine and hexuronic acid. Their generic sequence is [(1 ,4)-α-D-glucosaminyl-(1 ,4)-β-D- hexuronosyljn, with n = 50 to 150, in which the glucosamine is either N- acetylglucosamine (GlcNAc) or N-sulfoglucosamine (GlcNS) and the hexuronate is present as either glucuronate (GlcA) or its C-5 epimer iduronate (IdoA). O-sulfation, which occurs predominantly at C-2 of IdoA and of the glucosamine residues, but also rarely at C-2 of GlcA and C-3 of glucosamine, adds structural complexity to the chain. At each step only a fraction of potential substrates are modified, resulting in a very large structural diversity. In order to be able to analyse glycosaminoglycan fragments with the same sensitivity as obtained for N-glycans, we adapt our methodology as follows: (1) Fluorescent labeling of saccharides: Carbohydrate samples comprising less than a picomole of the carbohydrate to be sequenced are evaporated to dryness and preferably labeled with APTS by dissolving the sample in formamide containing 1M sodium cyanoborohydride or an appropriate concentration of any other suitable reducing agent containing 10 mM APTS or any other suitable concentration. Less preferably, APTS is substituted by another label of which the fluorescence can be detected using a laser. The labeled saccharide is separated from the excess free label by passing the sample through a bed of Sephadex G10 and eluting the glycans with 4 times 10 microliter of water. (2) Structure determination: All chemical cleavage steps and enzymatic digests are performed as fully described in Turnbull, JE et al. (1999) Proc. Natl. Acad. Sci. USA 96, 2698. (3) Analysis: due to the inherent charge of the GAG chains, separation on size is more important than in N-glycan analytical separations. Therefore, a higher gel concentration is generally more favorable and we use 12, 15 and 20% gels to achieve a good resolution. Sample preparation and all other electrophoresis conditions are as described for N-glycans herein.

8. Use of a meltable polyacrylamide matrix for SDS-PAGE in a procedure for N-glycan analysis on picomole amounts of glycoproteins in a mixture

The analysis of the glycan structures present on glycoproteins is a challenging problem encountered in many biopharmaceutical, glycobiological and proteomics laboratories. Purifying these proteins by classical chromatographic techniques such as hydrophobic interaction chromatography and ion exchange chromatography without disturbing the glycoform distribution is often difficult as the carbohydrate chains present on the protein can strongly influence the physical properties of the glycoprotein entity. Because glycan structure strongly influences the purification behavior of a glycoprotein, it is of definite advantage to have a technology that can give the glycan structural information without having to go through purification optimization. Typically, one would like to immunopurify a small quantity of the protein under study and get the glycan structural information from there. However, glycoprotein impurities will often still be present and could significantly alter the obtained structural information. SDS-PAGE separation of the post-immunopurification mixture can be a powerful second purification step. However, methods to obtain the glycans after SDS-PAGE separation are limited. One successful approach is the in-gel digestion with PNGase F, after which the N-glycans are eluted from the gel and analysed. However, this method makes use of Coomassie Brilliant Blue detection of the proteins and has not been successfully applied on the submicrogram scale. In the present example we use a commercially available meltable polyacrylamide matrix and present a methodology involving SDS-PAGE using this matrix, detection of the separated proteins with submicrogram sensitivity using Sypro-Orange fluorescent staining, melting the gel piece containing the glycoprotein of interest, deglycosylation after binding of the protein to Immobilon P in a 96-well plate and analysis of the N-glycans according to the invention.

The Protoprep II SDS-PAGE kit with meltable polyacrylamide formulation was obtained from National Diagnostics in beta-test version. The kit is now available commercially from this manufacturer. Sypro-Orange was obtained from Molecular Probes. Immobilon-P lined 96-well filterplates were purchased from Millipore. Recombinant human EPO was obtained from Roche. All other chemicals were analytical grade reagents from major suppliers.

Samples for SDS-PAGE were prepared according to well known procedures in the art. The gel casting and electrophoresis procedures were performed according to the directions of the manufacturer. After completion of the separation, the gel was stained with sypro-orange (diluted 1 :5000 in 7% acetic acid) for 60 minutes and rinsed in 7.5% acetic acid for 30 seconds. The protein bands were visualized using UV- transillumination, as is normally done for ethidiumbromide-stained agarose gels in DNA-analytical procedures. The bands of interest were cut out and the gel pieces transferred to clean pre-weighed eppendorf tubes. The pH of the gel piece was adjusted using repetitive incubations at room temperature with a volume of the gel dissolution buffer sufficient for completely immersing the gel piece until the pH of the solution reached 9. The gel was dissolved using three volumes of the dissolution buffer and incubated at 60°C until complete dissolution was observed. The dissolved acrylamide was removed by acetonitrile precipitation. Batches of 100 μl acetonitrile were added to the sample until no further precipitation is observed. The resulting white-coloured mass was removed using centrifugation and the supernatant was recovered. Subsequently, 2.5 volumes of RCM buffer (8 M ureum, 360 mM Tris, pH 8,6 and 3,2 mM EDTA) were added to the sample, incubated at 50 °C for 1 h and processed entirely as previously described (Callewaert N et al (2001) Glycobiology 11(4)275). The procedure involves binding of the protein to an Immobilon P membrane in a 96-well plate, washing and deglycosylation of PNGase F. Subsequently, the glycans are derivatized with APTS (8-aminopyrene-1 ,3,6-trisulfonic acid) and the excess APTS is removed over a Sephadex G10 bed packed in another 96-well filterplate. Following this, the labeled glycans are analysed on a 10% polyacrylamide gel on a standard Applied Biosystems 377A DNA-sequencer. The detection sensitivity of this technology is about 3 fmol of labeled carbohydrate. To test the performance of the developed procedure, we analysed a dilution series of recombinant human EPO. 1 unit of rhuEPO equals 10-15 ng of the protein. The lowest readily detectable amount of EPO on a standard UV transilluminator after Sypro Orange staining of the gel was 5 units (50-75 ng, 1-2 pmol). The bands, corresponding to 100, 50, 10 and 5 units of EPO were cut out, as well as a piece of a lane where no protein was loaded, as a blanc control. The result of the N-glycan analysis of these samples shows that the profiles obtained from the larger amounts of EPO are exactly reproduced by the profiles obtained from the smallest amount and very satisfactory profiling is possible with this tiny amount of protein. To our knowledge, this is the smallest amount of glycoprotein of which the glycans have ever been reported to be successfully analysed after SDS-PAGE purification of the protein. This level of sensitivity can also have utility in proteomics applications after 2D electrophoretic separation of a mixture of proteins. Materials and Methods 96-well deglycosylation procedure

The protocol was elaborated in detail by Papac et al.⁴. Briefly, the PVDF membrane at the bottom of the wells of a Multiscreen-IP plate (Millipore, Bedford, CA, USA) was wetted with 100 μl methanol, washed three times with 300 μl of water and once with 50 μl of RCM buffer (8M urea, 360 mM Tris, pH 8.6, 3.2 mM EDTA). The glycoprotein was loaded in the wells, containing 10 μl RCM buffer. Subsequently, additional RCM buffer was added to a minimal volume of 50 μl. Protein was bound to the membrane with a gentle vacuum. This step was followed by two wash steps with 50 μl RCM buffer. The bound protein was then reduced by the addition of 50 μl of 0.1 M dithiothreitol in RCM buffer and incubation at 37°C for 1h. The reducing solution was removed by vacuum and the wells were washed three times with 300 μl of water. Carboxymethylation was performed by addition of 50 μl of 0.1 M iodoacetic acid in RCM buffer and incubation for 30 min at room temperature in the dark. After removal of this solution, three washes with 300 μl of water followed. The remaining protein binding capacity of the wells was blocked by incubation with 100 μl 1 % polyvinylpyrrolidone 360 in water at room temperature for 1 h. Again three washing steps as described above were performed, followed by the addition of 1.25 Oxford Glycosystems units of PNGaseF (Oxford Glycosystems, Abingdon, UK) in 20 μl of 10 mM Tris-acetate pH 8.3. Digestion is complete after a 3h incubation at 37 °C, after which the solution is transferred by multichannel pipette to a tapered-well microtiter plate.

96-well AG-50-WX8 cation exchange If MALDI-TOF-MS of the analytes is required, a fraction of the deglycosylation mixture is treated with 150 mM acetic acid for 3h at room temperature to assure complete conversion of glycsosylamines to the reducing saccharides. Subsequently, this mixture is applied to the wells of a Multiscreen-Durapore membrane-lined 96-well plate (Millipore, Bedford, CA, USA), filled with 300 μl of AG-50-WX8 resin in the proton form (Biorad, Hercules, CA, USA). Plates are packed using the 100μl Multiscreen Column Loader system (Millipore, Bedford, CA, USA). In two rounds of resin loading, swelling and gentle centrifugation, microcolumns of about 300 μl packed resin are easily and reproducibly obtained. The cation exchange resin removes the protein and salt present in the deglycosyation mixtures with sufficient efficiency to allow direct MALDI- TOF-MS as described elsewhere⁴ (results not shown).

APTS derivatisation reaction We have found it unnecessary to remove the PNGase prior to derivatisation with APTS, as this practice does not lead to the appearance of contaminant peaks in the size range of 3-25 glucose units. The deglycosylation mixture was evaporated to dryness at the bottom of the tapered well microtiterplate using a Savant vacuum centrifuge equipped for plates and a 1μl 1 :1 mixture of 20 mM APTS (Molecular Probes, Eugene, CA, USA) in 30% acetic acid and 1M NaCNBH₃ in DMSO was added to each well. After carefull vortexing and short centrifugation of the plate, it was incubated up side down at 37 °C overnight, tightly wrapped in parafilm. The following morning, the reaction is quenched by the addition of 20 μl of water.

96-well Sephadex G10 post-derivatisation cleanup

The wells of a Multiscreen-Durapore membrane-lined 96-well plate (Millipore, Bedford, CA, USA), were packed as described for the AG-50-WX8 plates, but now with a swollen volume of 300 μl Sephadex G10. It is essential that the wells be centrifuged to dryness just prior to sample loading. After loading, the resin beds are eluted twice by addition of 25 μl of water and a 10 second centrifugation at 2000 rpm in a table-top centrifuge. The eluate is collected in another tapered-well microtiterplate and evaporated to dryness. Succesfull cleanup is hallmarked by the detection of only faint fluorescence of the eluate upon imaging on a standard UV-light box. After evaporation, the glycans are reconstituted in 5 μl of water.

Exoglycosidase digestions

0.8 μl batches are transferred to 250 μl PCR tubes or tapered well microtiterplates for treatment with exoglycosidase arrays. In this study, all digestions were done by overnight incubation at 37°C in 10 μl 20 mM sodium acetate pH 5.5 containing following enzyme mixtures; 1) Arthrobacter ureafaciens sialidase (2 U/ml, Boehringer, Mannheim, Germany); 2) sialidase and Diplococcus pneumoniae β-1 ,4-galactosidase (1 U/ml, Boehringer, Mannheim, Germany); 3) sialidase, galactosidase and Jack bean β-N-acetylhexosaminidase (30 U/ml, Oxford Glycosytems, Abingdon, UK) and 4) sialidase, galactosidase, N-acetylhexoaminidase and Jack Bean α-mannosidase (100U/ml, Sigma Biochemicals, Bornem, Belgium).

Preparation of the samples for gel loading To each sample, 0.5 μl of the ROX-labelled Genescan™ 500 standard mixture (Perkin Elmer, Foster City, CA, USA) is added, (alternatively, a mixture of ROX-labelled 6-, 18- ,30-,42-meric oligonucleotides (Life Technologies, Merelbeke, Belgium), each at a concentration of 250 fmol/μl can be used, concentrations calculated from the manufacturer's data), followed by 1 μl deionized formamide to facilitate sample loading.

Gel electrophoresis and data analysis

All experiments were performed on a Applied Biosystems 377A DNA-sequencer

(Perkin Elmer, Foster City, CA, USA), equipped with an external cooling bath (model RTE 111 , NESLAB, Porthsmouth, NH, USA) kept at 23°C (easily connectable to the sequencer according to the ABI PRISM 377 DNA sequencer user bulletin 'Modifications for subambient temperature operations'). Due to the high diffusional mobility of carbohydrates, we skip one gel lane between each two samples, in order to avoid cross-contamination and ease the lane tracking process. In the 36-well sequencing format, this allows analysis of 18 samples per run, and in the 64-well format, 32 samples can be analysed in parallel. The gel contains 12% of a 19:1 mixture of acrylamide:bisacrylamide (Biorad, Hercules, CA, USA) and is made up in the standard DNA-sequencing buffer (89 mM Tris, 89 mM borate, 2.2 mM EDTA). Polymerization is catalyzed by the addition of 200 μl 10% ammoniumpersulfate solution in water and 20 μl TEMED. The gels were of the standard 36 cm well-to-read length throughout the study. Prerunning is done at 3000 V for 1h. After prerunning the gel, the wells are thoroughly rinsed with the sequencing buffer and 1.8 μl of the samples is loaded. The electrophoresis voltage during separation is 4000 V and data are collected for 5h (separation of glycans up to 25 glucose units in size). Data analysis is performed using Genescan 3.1 software. We use the same fluorescence- overlap correction matrix as for DNA sequencing using BigDye dye terminators on our machine. The fluorescence of APTS-derivatised carbohydrates and rhodamine- labelled oligonucleotides is obviously readily resolved. References

1. Yoshimi, Y., Yamazaki, S. & Ikekita, M. Developmental changes in Asn-linked neutral oligosaccharides in murine cerebrum. Biochim Biophys Ada 1426, 69-79 (1999).

2. Dennis, J.W., Granovsky, M. & Warren, C.E. Protein glycosylation in development and disease. Bioessays 21, 412-21 (1999).

3. Cumming, D.A. Glycosylation of recombinant protein therapeutics: control and functional implications [see comments]. Glycobiology 1, 115-30 (1991). 4. Papac, D.I., Briggs, J.B., Chin, E. T. & Jones, A.J. A high-throughput microscale method to release N-linked oligosaccharides from glycoproteins for matrix-assisted laser desorption/ionization time-of-flight mass spectrometric analysis. Glycobiology 8, 445-54 (1998).

5. Chen, F. T. & Evangelista, R.A. Profiling glycoprotein N-linked oligosaccharide by capillary electrophoresis. Electrophoresis 19, 2639-44 (1998).

6. Guttman, A., Chen, F. T. & Evangelista, R.A. Separation of 1-aminopyrene- 3,6,8-trisulfonate-labeled asparagine- linked fetuin glycans by capillary gel electrophoresis. Electrophoresis 17, 412-7 (1996).

7. Turnbull, J.E., Hopwood, J.J. & Gallagher, J.T. A strategy for rapid sequencing of heparan sulfate and heparin saccharides. Proc Natl Acad Sci U S A 96, 2698-703

(1999).

8. Guile, G.R., Rudd, P.M., Wing, D.R., Prime, S.B. & Dwek, R.A. A rapid high- resolution high-performance liquid chromatographic method for separating glycan mixtures and analyzing oligosaccharide profiles. Anal Biochem 240, 210-26 (1996). 9. Suzuki, S. & Honda, S. A tabulated review of capillary electrophoresis of carbohydrates. Electrophoresis 19, 2539-60 (1998).

10. Rudd, P.M. & Dwek, R.A. Rapid, sensitive sequencing of oligosaccharides from glycoproteins. Curr Opin Biotechnol 8, 488-97 (1997).

11. Liu, J.P., Shirota, O., Wiesler, D. & Novotny, M. Ultrasensitive fluoromethc detection of carbohydrates as derivatives in mixtures separated by capillary electrophoresis. Proc Natl Acad Sci U S A 88, 2302-6 (1991).

12. Morell, M.K., Samuel, M.S. & O'Shea, M.G. Analysis of starch structure using fluorophore-assisted carbohydrate electrophoresis. Electrophoresis 19, 2603-11 (1998). 13. Chen, F. T. & Evangelista, R.A. Analysis of mono- and oligosaccharide isomers derivatized with 9- aminopyrene-1 ,4,6-trisulfonate by capillary electrophoresis with laser- induced fluorescence. Anal Biochem 230, 273-80 (1995).

14. Evangelista, R.A., Guttman, A. & Chen, F. T. Acid-catalyzed reductive amination of aldoses with 8-aminopyrene-1 ,3,6- trisulfonate. Electrophoresis 17, 347-51 (1996).

15. Mort, A.J., Zhan, D. & Rodriguez, V. Use of scavenger beads to remove excess labeling reagents from capillary zone electrophoresis samples. Electrophoresis 19, 2129-32 (1998).

16. Kotani, N. & Takasaki, S. Analysis of 2-aminobenzamide-labeled oligosaccharides by high-pH anion- exchange chromatography with fluorometric detection. Anal Biochem 264, 66-73 (1998).

17. Colangelo, J. & Orlando, R. On-target exoglycosidase digestions/MALDI-MS for determining the primary structures of carbohydrate chains. Anal Chem 71, 1479-82 (1999). 18. Kuster, B., Wheeler, S.F., Hunter, A.P., Dwek, R.A. & Harvey, D.J. Sequencing of N-linked oligosaccharides directly from protein gels: in- gel deglycosylation followed by matrix-assisted laser desorption/ionization mass spectrometry and normal-phase high- performance liquid chromatography. Anal Biochem 250, 82-101 (1997).

19. Kuster, B., Hunter, A.P., Wheeler, S.F., Dwek, R.A. & Harvey, D.J. Structural determination of N-linked carbohydrates by matrix-assisted laser desorption/ionization- mass spectrometry following enzymatic release within sodium dodecyl sulphate- polyacrylamide electrophoresis gels: application to species-specific glycosylation of alphal -acid glycoprotein. Electrophoresis 19, 1950-9 (1998).

20. Guttman, A., Chen, F.T., Evangelista, R.A. & Cooke, N. High-resolution capillary gel electrophoresis of reducing oligosaccharides labeled with 1-aminopyrene-3,6,8- trisulfonate. Anal Biochem 233, 234-42 (1996).

Claims

Claims

1. An electrophoretic method to analyse underivatized carbohydrates present in less than 1 picomole amount in a sample.

2. An electrophoretic method according to claim 1 comprising: derivatizing said carbohydrates with a derivatizing agent, removing the portion of the derivatizing agent which did not derivatize with said carbohydrate,

- separating said derivatized carbohydrates, and - detecting said derivatized carbohydrates.

3. A method according to claim 2 wherein said removal of the portion of the derivatizing agent which did not derivatize with said carbohydrate is done with a gel filtration and/or gel permeation resin and/or size-selective membrane.

4. A method according to claim 3 wherein said resin is Sephadex G10.

5. A method according to claims 3-4 wherein said resin is packed in columns in which the flow is driven by centrifugal force or by a vacuum.

6. A method according to claim 5 in which said columns are incorporated in a multi-well format such as a 96-well plate.

7. A method according to any of claims 2 to 6 wherein the derivatizing agent is 8- amino-1 ,3,6-pyrenetrisulfonic acid.

8. A method according to any of claims 2 to 7 wherein said separation of said derivatized carbohydrates is performed with capillary electrophoresis and/or DNA- sequencing-equipment.

9. A method according to any of claims 2 to 8 wherein said detection is undertaken via a spectrometric and/or fluorometric detection system.

10. A device for the detection of less than 1 picomole amount of underivatized carbohydrates comprising: a derivatization chamber, a clean-up apparatus wherein said clean-up occurs via gel filtration and/or gel permeation resin which remove the excess fluorescent and/or spectrometric label, and an electrophoretic apparatus such as a capillary electrophoresis and/or DNA-sequencing equipment for the separation and detection of derivatized carbohydrates.

11. Use of the method according to claims 1-10 to obtain carbohydrate structural information.

12. Use of the method according to claims 1-10 for diagnosing diseases where at least one modification of one or more carbohydrates is implicated.

13. Use of the method according to claims 1-10 for the identification of carbohydrate structures of recombinant glycoproteins in biological fluids.

14. Use according to claim 11 wherein said glycoprotein is erythropoetin.

15. Use of the method according to claims 1-10 for the identification and/or structural characterization of carbohydrates which are bound to other biomolecules.

16. Use according to claim 15 wherein said carbohydrates are derived from organisms such as prions, viruses, mycoplasma, bacteria, fungi or parasites.